Tuesday, February 23, 2010

How Do I Determine Estimates of the Size Attributes?

CMMI-DEV Project Planning (PP) Specific Practice (SP) 1.2 states 'Establish and maintain estimates of the attributes of the work products and tasks.'

During appraisals it has been noted that most of the time we are relying on expert knowledge to determine effort and cost rather than estimating the sizing attributes. Just using expert knowledge makes it difficult to improve our effort estimation, hence having a dedicated set of attributes is essential in order for us to improve our estimations. We have evaluated some comprehensive methods (e.g. Function Points), but these are not considered as beneficial to our projects. We are now looking for an approach that allows us to improve our estimates with a simpler method which could then be refined based on our needs. Would you please give me advice how to tackle this challenge?

What PP SP 1.2 is looking for is the Basis of Estimate for the work product. If you are estimating a document for example, what aspects or attributes of the document can you estimate that when combined with a productivity factor will yield effort and cost? You determine these attributes by analyzing the historical data from your organization’s past projects. So based on the project’s requirements, you could estimate the number of pages that would be written for a document, the number of figures or drawings that need to be created or modified, etc. These items are then sizing parameters.

For source code, you need to analyze historical data from past projects to give an empirical estimate. As you say, function points are comprehensive. A simpler method may be to classify your software requirements in categories such as interface requirements, display requirements, processing requirements, reporting requirements, etc. and then using your historical data determine the correlations between these requirements categories and code size. Many organizations start with this simple kind of estimation/prediction model and continually refine it by incorporating data from each new project until a fairly accurate estimation model emerges. This estimation model is then very specific to the organization and type of software development it performs.

Hope these ideas help.

No comments: